The Novula Lens — Our Thesis

The Map · The AI economy is a stack

AI runs on atoms.

Long before a model writes a single sentence, something has to generate the electricity, forge the chips, build the factory, and — increasingly — move the robot. Five floors, one building.

5

Action The hands

Software that acts — and now, machines that act in the physical world.

4

Intelligence The minds

The models themselves. The floor everyone photographs.

3

Infrastructure The factories

Vast data centers where the engines are racked, cooled, and run day and night.

2

Silicon The engines

The chips that do the thinking — and the memory that feeds them.

1

Power The foundation

AI is the most electricity-hungry technology ever built. Nothing above runs without it.

You notice the top floor — but the building is only ever as tall as its foundation will bear.

The Thesis · The center of gravity is descending

Atoms versus bits.

For a decade the bottleneck kept changing floors — algorithms, then chips. It's still sinking. The new scarcity isn't intelligence; it's electricity, and the machinery that turns intelligence into action. That's a different contest entirely.

BITSthe top of the stack

Intangible, fast, copied for free
Winner-take-most
Rewards talent, venture capital & moonshots

America's home game — and America is very good at it

ATOMSthe bottom of the stack

Power plants, factories, batteries, motors
Slow, capital-heavy, compounds for decades
Rewards patience & building capacity before demand

China's home game — and nowhere plays it better than the GBA

The real "why China" isn't a slide of statistics. Every floor the AI economy descends moves onto ground China spent twenty years building while the rest of the world wasn't looking.

Four reasons it compounds rather than fades.

01

Cost falls with volume, and supply chains cluster

The more you build, the cheaper each one gets — and the know-how pulls together in one place. It's why China's lead in solar, EVs and drones became near-total, not marginal.

02

You can't rent a decade of power

You can rent compute, license a model, or source chips. You cannot rent twenty years of grid overbuild.

03

Every restriction accelerates its target

Each export control speeds China's drive to build its own — chips, memory, robot components. The wall makes the muscle.

04

An aging society is a built-in customer for machines

A shrinking workforce gives China a reason to deploy physical AI at a scale few other places share.

We hold this honestly — which means holding its rebuttal too

America still owns the two things that may matter most: the intelligence ceiling, and the capital funding the entire build-out. The serious counter-argument is that intelligence is the master lever — that enough of it can solve the atoms problems. We take that seriously. So our claim is deliberately narrow. We do not say China wins AI. We say the world is fixated on the model floor and underpricing the physical floors it is about to depend on — and those floors are China's, and most densely the Greater Bay Area's. That gap between attention and reality is the opportunity. Closing it is what we do.

And this is not a story about America stumbling. It's a patient, decades-long bet on the physical foundation of technology, vindicated by a wave nobody could have named in advance. China didn't out-sprint the frontier. It built the ground floor — and the whole industry is now discovering it needs one.

The Proof · Physical AI

If robots are the next frontier, the frontier has an address.

Nvidia's Jensen Huang calls physical AI — intelligence that leaves the screen and enters the world — the next frontier, a market he sizes in the trillions. Take him at his word and ask the obvious question: where on earth should you stand to catch that wave?

~4 in 5

humanoid robots made worldwide ship from China — at a fraction of Western cost

Atoms

a robot is mostly atoms — motors, joints, batteries, frame. The body is the majority of the cost

1 belt

the Shenzhen–Dongguan hardware belt is the densest, fastest supply chain for that body on earth

1 border

between that factory floor and Hong Kong's global capital market — nowhere else on the planet

America builds the brain.
The Bay Area builds the body.

When Nvidia unveiled its reference humanoid, it paired American compute with a Chinese-made body · The Novula Lens

A humanoid robot working on an EV production line at XPENG — On the line at XPENG — from our field notes. This supply chain is the ground we walk clients across: DJI, UBTECH, XPENG, Pony.ai.

If physical AI is the next frontier, the GBA isn't near the answer. It is the answer.

One honest note, because credibility matters more than hype: today's robots are early — the best industrial humanoids still work at a fraction of human efficiency, and China dominates the manufacturing base, not yet the capability frontier. But manufacturing dominance is exactly the advantage that compounds. Ask solar. Ask EVs. Ask drones.

The Vehicle · The Novula methodology

A worldview is only worth having if you can act on it.

Three tiers, one methodology. The Map — how the AI economy is built. The Thesis — its center of gravity is descending to the physical, and the physical is the GBA's. The Vehicle — the four ways we take you there.

01

Innovation Tours

Come and stand at the center of gravity.

Step inside the factories, labs and supply chain the thesis describes. Explore the tours →

02

Bootcamp

Learn to read the stack for yourself.

A selective, live-in program where the worldview becomes working knowledge. See the bootcamp →

03

Incubator

Build something on it.

For founders ready to act on the thesis — from prototype to scale, on the ground floor itself. See the incubator →

04

Invest

Allocate to the floors the world is mispricing.

For a small circle of partners, we turn the gap between attention and reality into positions. Investor access →

Underneath all four runs the same method — Understand → Curate → Immerse → Connect. Map → Thesis → Vehicle is the why; the Method is the how.

The entry point

Come stand at the center of gravity.

The thesis takes a page to read and a week to believe. Start with a tour.

Explore the tours See the numbers first

Written as of mid-2026. Humanoid shipment share per Omdia / Bloomberg (Chinese firms shipped the large majority of the ~13,000 humanoids delivered in 2025); "physical AI" framing and market sizing attributed to Nvidia CEO Jensen Huang (CES & GTC keynotes); Nvidia's Isaac GR00T reference humanoid pairs Nvidia compute with a Unitree body. Figures in this space move monthly — we update this page as they do.